← all workloads

control_flow_ssa

A single entry point with n stacked control-flow blocks — nested if/else, a bounded inner loop with break/continue, and a switch — all mutating a small set of carried locals (a, b, c, d). Re-assigning the same locals across many branches and loop back-edges is what forces SSA construction to insert phi nodes and what the CFG simplifier (constructSSA, inside simplifyIR) must chew through. Isolates the SSA/CFG axis that complexity_ladder only touches as one of several mixed dimensions; nothing else stresses it alone. Scales by breadth = number of control-flow blocks, so basic-block and phi counts grow with n. Scaling null: n scales control-flow blocks; ideal SSA construction is near-linear in block count.

bucket: control_flow  ·  mode: target  ·  flags: -target spirv -emit-spirv-directly

Phase composition vs N (stacked sub-counters)

compileInner split into phase buckets (named leaves + (self) residuals) stacked across the sweep sizes — the top edge is compileInner, so you can see which phase drives the scaling.

control_flow_ssa — phase composition vs N (v2026.5, median ms) control_flow_ssa 13.2× over N 60→480 0.0 899 1798 60 120 240 480 N control_flow_ssa — parseTranslationUnit control_flow_ssa — SemanticChecking control_flow_ssa — generateIR control_flow_ssa — frontEndExecute (self) control_flow_ssa — specializeModule control_flow_ssa — simplifyIR control_flow_ssa — linkIR control_flow_ssa — unrollLoopsInModule control_flow_ssa — legalizeResourceTypes control_flow_ssa — legalizeExistentialTypeLayout control_flow_ssa — performMandatoryEarlyInlining control_flow_ssa — performForceInlining control_flow_ssa — linkAndOptimizeIR (self) control_flow_ssa — generateOutput (self) control_flow_ssa — compileInner (self) phase buckets parseTranslationUnit SemanticChecking generateIR frontEndExecute (self) specializeModule simplifyIR linkIR unrollLoopsInModule legalizeResourceTypes legalizeExistentialTypeLayout performMandatoryEarlyInlining performForceInlining linkAndOptimizeIR (self) emitEntryPointsSourceFromIR generateOutput (self) compileInner (self)

Scaling analysis

floor-subtracted power-law fit (t − floor) = a·Nk; floor = the minimal workload (fixed per-compile cost), k the global exponent, top-2× the local high-end doubling ratio.

N rangefloor (ms)k (work)fit R²t(Nmin)t(Nmax)top-2×
60–48091.280.99312616652.83×

Growth attribution (N=60 → N=480)

compileInner grows by 1539 ms across the sweep; the mutually-exclusive phase buckets below partition that growth exactly (no nested-timer double counting). × lin is the same metric as the top-level panels, per bucket: the end point vs a linear expectation anchored to the bucket's share of the minimal floor and fitted on the low-N half — 1.0 = grew exactly linearly, >1 bends up. The super-linearity lives where × lin (and k) are red.

buckett@N=60t@N=480Δ msshare× lin∝Nk
generateOutput (self)31567+53635%2.05×1.43
linkAndOptimizeIR (self)13526+51333%3.39×1.79
generateIR13160+14710%1.62×1.22
simplifyIR13125+1117%1.17×1.08
SemanticChecking46147+1017%0.60×0.59

Also growing (below top-5): performMandatoryEarlyInlining (+78 ms, 5%).

Near-constant (≤2% of growth each): specializeModule (3→25 ms), legalizeExistentialTypeLayout (1→11 ms), legalizeResourceTypes (1→11 ms), linkIR (1→7 ms), parseTranslationUnit (1→5 ms), performForceInlining (0→2 ms), unrollLoopsInModule (0→1 ms), frontEndExecute (self) (0→0 ms), compileInner (self) (0→0 ms).

Sweep numbers (median ms)

NcompileInnersimplifyIRfrontEndExecute
601261360
1202462786
24058957147
4801665125312